Abstract

The recent trend for analyzing 3D shapes in medical application has arisen new challenges for
a vast amount of research activities. Quantitative shape comparison is a fundamental problem
in computer vision, geometry processing and medical imaging. This thesis is motivated by the
availability of carpal bone shape dataset to develop efficient techniques for diagnosis of a variety
of wrist diseases and examine human skeletal.
This study is conducted in two sections. First, we propose a spectral graph wavelet approach
for shape analysis of carpal bones of the human wrist. More precisely, we employ spectral graph
wavelets to represent the cortical surface of a carpal bone via the spectral geometric analysis of the
Laplace-Beltrami operator in the discrete domain. We then propose global spectral graph wavelet
(GSGW) descriptor that is isometric invariant, efficient to compute and combines the advantages of
both low-pass and band-pass filters. Subsequently, we perform experiments on shapes of the carpal
bones of ten women and ten men from a publicly-available database of wrist bones. Using one-way
multivariate analysis of variance (MANOVA) and permutation testing, we show through extensive
experiments that the proposed GSGW framework gives a much better performance compared to
the global point signature (GPS) embedding approach for comparing shapes of the carpal bones
across populations.
In the second section, we evaluate bone age to assess children’s biological maturity and to diagnose
any growth disorders in children. Manual bone age assessment (BAA) methods are timeconsuming
and prone to observer variability by even expert radiologists. These drawbacks motivate
us for proposing an accurate computerized BAA method based on human wrist bones X-ray images.
We also investigate automated BAA methods using state-of-the-art deep learning models that
estimate the bone age more accurate than the manual methods by eliminating human observation
variations. The presented approaches provide faster assessment process and cost reduction in the
hospitals/clinics. The accuracy of our experiments is evaluated using mean absolute error (MAE),
and the results demonstrate that exploiting InceptionResNet-V2 model in our architecture achieves
higher performance compared to the other used pre-trained models.